Nerf repeated angles in Flashlight skill

[MBmasher]

Originally this PR aimed to nerf grid style mapping, however the method used to do it wasn't ideal, and it would take a lot of work to properly consider the difficulty of grid style mapping.

Now it simply nerfs repeated angles, which are generally easy to read and make the placement of objects predictable. This indirectly nerfs grid style maps.

Nerf examples:
Night of Knights [SOLO] (DJPop) +HDHRFL: 837pp -> 689pp
Kakenukeru Anime Song Medley [TWO DIMENSIONS] (Monstrata) +FL: 847pp -> 753pp
Cross Over [Ultra] (Sonnyc) +FL: 547pp -> 510pp

---

Spreadsheet of SR/PP changes: https://docs.google.com/spreadsheets/d/1TKGEQSsG0-Wbkz6bRi_6hOW2fg5uxXxa2xXDIexRlk4/edit

As of b082dc1

[smoogipoo]

SR/PP sheet added to OP.

[stanriders]

Can't say I'm a big fan of the way its done, I think it'd be better to pre-pass the map to calculate angles and nerf angles that are being repeated too much (that would also nerf other non-grid but easy to memorize maps) or even go through the map and see if it lands on the grid and apply a nerf based on that.

The way its currently done is viable too tho, if you feel like proper realization would take too much time/effort

[MBmasher]

I personally agree @stanriders however this is probably fine for the next deploy at least.

Looking at the spreadsheet, can't say there's any values that I'm unhappy to see, it's basically what I expected

[apollo-dw]

If you want to target only patterns within a grid, then I think it'd be a good idea to check whether the Previous object and maybe even the PreviousPrevious object have those angles too. Otherwise I feel like we're nerfing any pattern with those angles.

[MBmasher]

That's a good idea @apollo-dw, and probably worth trying. However practically this on its own seems to be fine; normal patterns mapped by hand rarely ever use these specific angles, let alone anywhere near as often as grid maps do, so they aren't affected too much... that's just a guess on my part though

[Flutterish]

I agree with @stanriders , and it should be fairly easy to do with a simple Fourier transform to determine the frequency of each angle. It would be a better approach than for example angle buckets where we just count the amount of each angle, as the frequency curve can better match slightly different angles, or angles at the edge of a bucket range as belonging to a given neighborhood.

[apollo-dw]

@smoogipoo Needs sheet update

[smoogipoo]

@apollo-dw Sheet updated.

[stanriders]

lgtm

[apollo-dw]

👍